Orchestration Specialist

Full source: docs/ai-experts/02-orchestration-specialist.md

Scope

Astronomer Airflow DAG development (TaskFlow API), Azure service integration in DAGs, and keeping the utils wheel correctly bundled for the Airflow environment.

Key constraints

• orchestration/ has its own UV workspace, isolated uv.lock, separate from the root.
• utils is bundled as a pre-built wheel (include/wheels/utils-<version>-py3-none-any.whl), not a workspace reference — rebuild via make bin after any utils change.
• DAGs use the TaskFlow API (@task decorators) and Asset-based scheduling for event-driven triggers (e.g. AsbQueueEventTrigger for Azure Service Bus messages).
• Secrets are fetched lazily at task execution time, never at DAG parse time.

Build workflow

cd orchestration
make bin      # rebuild utils wheel + relock
make build    # Docker image w/ BuildKit secrets
make start    # local Astro dev

When to consult

• Creating new Airflow DAGs or custom sensors/triggers
• Debugging Airflow task execution environments
• Updating utils and rebuilding wheels
• Deploying to Astronomer Cloud

See also: Orchestration for the current DAG inventory and deploy pipeline, Packages → db for the database package DAGs depend on.